The volume product of convex bodies with many hyperplane symmetries Franck Barthe, Matthieu Fradelizi To cite this version: Franck Barthe, Matthieu Fradelizi. The volume product of convex bodies with many hyperplane symmetries. American Journal of Mathematics, Johns Hopkins University Press, 2013, 135 (2), pp.1- 37. 10.1353/ajm.2013.0018. hal-00793781 HAL Id: hal-00793781 https://hal-upec-upem.archives-ouvertes.fr/hal-00793781 Submitted on 22 Feb 2013 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. The volume product of convex bodies with many hyperplane symmetries F. Barthe and M. Fradelizi To appear in Amer. J. Math. Abstract Mahler's conjecture predicts a sharp lower bound on the volume of the polar of a convex body in terms of its volume. We confirm the conjecture for convex bodies with many hyperplane symmetries in the following sense: their hyperplanes of symmetries have a one-point intersection. Moreover, we obtain improved sharp lower bounds for classes of convex bodies which are invariant by certain reflection groups, namely direct products of the isometry groups of regular polytopes. Mathematic Subject Classification: 52A20; 52A40; 52A38; 52B15; 46B10 1 Introduction We work in Rn equipped with its canonical Euclidean structure. The scalar product is denoted by h·; ·i and the norm by j · j. Let K be a convex body in Rn. For z 2 Rn, the polar body of K with respect to z is defined by z n K = y 2 R ; hy − z; x − zi ≤ 1 for every x 2 K : We write K◦ for z = 0, which should not be confused with Int(K), the interior of K. In what follows, the n volume of a convex set K ⊂ R is denoted by jKjn and simply jKj when there is no ambiguity. The volume product of K is by definition P(K) := inf jKj jKzj: n z2R It is known (see [15]) that the above infimum is achieved for a unique point, called the Santal´opoint of K and denoted by s(K). The volume product and the Santal´opoint are affine-invariant in the sense that for every non singular affine transform A, one has P(AK) = P(K) and s(AK) = As(K). Hence if there are affine maps which fix K globally and whose only common fixed point is the origin then s(K) = 0 and P(K) = jKj jK◦j. This is the case in particular if K is origin-symmetric (i.e. K = −K). A celebrated result of Blaschke ([1] for n = 2 or 3) and Santal´o([24] for general n) asserts that among n-dimensional convex bodies, the Euclidean ball has maximal volume product. The so-called Mahler's conjecture (see [12]) asserts that, among convex bodies having a center of symmetry, the volume product should be minimal for the cube (and also for its dual and for some combinations of the two sets, see the end of section 5 for more details), equivalently, the conjecture predicts that for such a convex body K ⊂ Rn, 4n jKj jK◦j ≥ jBnj jBn j = ; 1 1 n! n n n n P p where B1 = [−1; 1] and for p 2 (0; 1), Bp = fx 2 R ; jxij ≤ 1g. A corresponding question asks if among general convex bodies the volume product is minimal for simplices: is it true that for every convex body K ⊂ Rn and every z 2 Rn the following inequality holds (n + 1)n+1 jKj jKzj ≥ j∆ j j(∆ )◦j = ; n n (n!)2 1 n where ∆n is a regular simplex in R , with 0 as center of mass. The conjectures were fully confirmed in dimension 2 by Mahler [13]. A few positive results were obtained in arbitrary dimension. The symmetric version of the conjecture was established for unconditional bodies by Saint Raymond [23] (see also Meyer [14] for a simple proof). Recall that a convex body in Rn is called unconditional if for all choices of signs " = ("1;:::;"n) 2 {−1; 1g we have x = (x1; : : : ; xn) 2 K =) ("1x1;:::;"nxn) 2 K: Equivalently, Si(K) = K for i = 1; : : : ; n where Si denotes the orthogonal symmetry with respect to the n coordinate hyperplane fx 2 R ; xi = 0g. The conjecture was proved by Reisner [21] for zonoids (see also [8] for a simple proof) and by Meyer and Reisner [17] for polytopes with less than n + 3 vertices. For convex bodies close, in Banach-Mazur distance, to the cube (resp. to the simplex) the inequalities were established by Nazarov, Petrov, Ryabogin and Zvavitch [19] (resp. by Kim and Reisner [10]). An asymptotic version of the conjectures was also considered. Bourgain and Milman [2] established that there is a universal constant c > 0 such that for every integer n ≥ 1 and every convex body K in Rn, 1 c P(K) n ≥ : n Other proofs were given by Pisier [20] and more recently by Kuperberg [11] and Nazarov [18]. Functional versions of Mahler's conjectures were also investigated by the second author and Meyer [5, 6, 7]. The aim of the present paper is to confirm Mahler's conjecture under rather general symmetry assump- tions and to improve the conjectured bound in certain cases. The idea of the proof is to adapt Meyer's argument [14] for unconditional convex bodies to the case of more general symmetry assumptions. To state our results more precisely, we need to introduce some notations. We denote the orthogonal group (the set of linear isometries of Rn) by O(Rn). For a subset A ⊂ Rn, we define the group of linear isometries of A as O(A) = fg 2 O(Rn); g(A) = Ag. Also span(A), conv(A) and pos(A) denote respectively the smallest vector space, the smallest convex set and the smallest convex cone containing A. Given a vector n n subspace E ⊂ R , PE stands for the orthogonal projection onto E and SE = 2PE − IdR for the orthogonal symmetry with respect to E. The orthogonal symmetry with respect to a hyperplane is also called a reflec- tion. Eventually, if a group G acts on Rn, Fix(G) = fx 2 Rn; 8g 2 G; gx = xg is the set of points which are fixed by the whole group. The main results of this article are contained in the following theorem. Theorem 1. Let n ≥ 2 and K ⊂ Rn be a convex body. (i) If O(K) contains O(P ) for some regular polytope P in Rn then P(K) ≥ P(P ) with equality if and only if K is a dilate of P or P ◦. n1 (ii) If O(K) contains O(P1) × · · · × O(Pk) for some regular polytopes or Euclidean balls P1 in R ,..., Pk n in R k , with n1 + ··· + nk = n then P(K) ≥ P(P1 × · · · × Pk). (iii) If O(K) contains m reflections SH1 ;:::;SHm with respect to m hyperplanes H1;:::;Hm such that Tm i=1 Hi = f0g then P(K) ≥ P(∆n); with equality if and only if K is a simplex. Section 2 is devoted to convex sets in the plane. We establish the sharp lower bounds under symmetry assumptions and prove that regular polygons play an extremal role. This study allows to put forward some of the key ingredients which will be crucial in higher dimensions. In particular, the method requires an S essential partition of the convex body K of the form K = i K \ Ci where the sets Ci are convex cones with the property that any point y 2 Ci is \normed" by a point of K \ Ci in the following sense sup hy; xi = suphy; xi: x2K\Ci x2K 2 Section 3 shows how to ensure this property when a convex body K in arbitrary dimension has many symmetries. Section 4 deals with regular polytopes. The main result there is that a convex body having the same symmetries as a regular polytope has a larger volume product. This is a warm-up for Section 5 where a similar statement is established for convex bodies which are invariant by the direct product of the isometry groups of regular polytopes. Section 6 deduces from the result of Section 5 that Mahler's conjecture is indeed true for convex bodies having enough hyperplane symmetries. The classification of reflection groups plays a crucial role here. In section 7, we deduce Mahler's conjecture for some bodies of revolution, equivalently we establish functional forms of Mahler's conjecture for concave functions having many symmetries. The final section collects remarks and further questions. 2 In the plane Given a convex body containing the origin it its interior, a natural strategy in order to evaluate jKj jK◦j is to divide the plane into convex cones with apex at the origin and to work separately in each cone. This is possible only when K meets the boundary of the cone in a certain way, as the next statement shows. 2 Proposition 2. Let α 2 (0; π) and u1; u2 be unit vectors in R with angle α, i.e. hu1; u2i = cos α. Denote + by C := pos(u1; u2) the positive convex cone generated by u1; u2. Let Ci := R ui and Pi the orthogonal projection onto Rui.
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages28 Page
-
File Size-